top of page

HOME  |  FILMS  |  REVIEWS

Gradient Descent

average rating is 3 out of 5

Critic:

Lawrence Bennie

|

Posted on:

Dec 23, 2025

Film Reviews
Gradient Descent
Directed by:
Vir Srinivas
Written by:
Vir Srinivas
Starring:
James Allen Barnes, Peter Mendes, Anjie Parker

During a recent trip to the Highlands, a moment stood out for me during a gondola trip up and down Aonach Mòr.  For late December, the weather was very good, allowing for a spectacular view of the Nevis Range, nearby Fort William and scenic Highland beauty during the descent to basecamp.  As I passed another carriage, I noticed that the individual seated inside was oblivious to the dramatic landscape on display around them.  The reason why is not probably going to be a surprise.  They were too engrossed on their phone.  Later, at a church service, the parish priest lamented about the account of a 12 year-old boy who was denied his phone for three hours during the filming of a TV documentary.  The result?  The boy wrecked his house.  

 

Such anecdotes may well induce raised eye-brows or groans from sceptical readers, but such warning signs of alienation, disconnection and isolation are at the heart of writer-director Vir Srinivas' Gradient Descent.  However, this descent is not a physical one down from a mountain but a psychological drop from desperation to desensitization and Srinivas' dramatic conflict is not just simply the mobile phone dilemma but a greater danger: AI.

 

Jackson (James Allen Barnes) is down-and-out; unemployed, homeless and left to beg on the city streets.  Yet, salvation arrives in the form of a well-dressed stranger (Peter Mendes) who offers Jackson an easy-fix job.  Before long, Jackson is suited and booted, attending first-rate AI training, moving back in with partner Maya (Anjie Parker) and working away in a world of waiting and watching as he and his colleagues silently categorise on-going surveillance footage.  However, what begins as a dream soon descends into nightmarish territory as Jackson helplessly becomes the onlooker to increasing acts of dehumanizing brutality and, in the process, finds his own compassionate awareness disturbingly slipping away...

 

Srinivas opens Gradient Descent with Jackson barely noticeable, dwindled by the sheer size of the buildings and noise of the city around him.  He is truly down and out and even the viewer only notices him once the camera moves closer.  Expectedly, he is ignored by passers-by despite pleas for help.  Unexpectedly, he is approached by a slick-looking businessman who Jackson locks onto from a distance.  It's a decidedly Hitchcockian sequence from Srinivas cutting between Jackson and his subjective view of the stranger approaching and we're left nervingly unsure of what's unfolding.  Will Jackson attack the man, in a final act of desperation for survival?  Does he already know him?  Instead, a Squid Game-esque scenario plays out, with the stranger charming our vagrant protagonist into a shadowy enterprise that feels too good to be true.

 

Unfortunately, it is.  There's a trade off for the fixer's golden ticket.  In scenes which echo the likes of A Clockwork Orange and The Parallax View, Jackson is soon exposed to escalating unsettling content, the worst of which is actually kept from the viewer.  Soon enough, Jackson begins to disconnect, spurning intimacy from his love and ignoring the cries of his baby as we see him consumed by his new-found obsession; the mechanics of artificial intelligence.  "I won't let them get away with it", he says.  

But it's too late.  The seeds of disassociation have been planted and by the film's shocking end Jackson has become another lost soul in a sea of silent surveillance.  It's an arresting final sequence from Srinvas' which makes up for the film's clunkier elements (the villains are unconvincing and their central scene together plays on for too long) and which ends the piece on a powerfully conflicting note of both horror and hope.

Podcast Film Reviews
About the Film Critic
Lawrence Bennie
Lawrence Bennie
Short Film
bottom of page